Machine Learning’s Missing Link?

Machine Learning’s Missing Link

MLOps (a portmanteau of ‘machine learning’ and ‘operations’) refers to data science operationalization. It focuses on providing capabilities to roll out machine learning models across an organization after the models have been created.

The machine learning workflow

In order to benefit from machine learning technologies, organizations must support all of the stages in a machine learning workflow:

  1. Data collection
  2. Structuring
  3. Modeling
  4. Deployment into production environments

Many data science platforms skimp on the deployment piece, catering to all the capabilities required to create a model, but falling short when it comes to going live successfully. In this sense, MLOps is the ‘missing link’ when it comes to widespread machine learning adoption.

What’s driving machine learning adoption

The Current and Future State of AI and Machine Learning
The Current and Future State of AI and Machine Learning
Download report >>

Machine learning’s popularity has driven a flurry of startup activity, resulting in the emergence of about a dozen providers looking to support machine learning models and data-driven decision-making.

A recent 451 Alliance survey points to digital security as the most important business reason for employing machine learning (9.3% of respondents). Analytics use cases accounted for a far greater percentage when considered cumulatively.

Altogether, 22.1% of enterprises said analytical use cases, such as marketing, logistics and inventory analysis, were the most important reason for using machine learning.

Traditionally, operationalization hasn’t been a top priority for machine learning vendors. However, with wider adoption across individual organizations and across industries, MLOps is becoming increasingly important.

MLOps Tools

Some vendors are developing their own MLOps tools for their platforms, while others are acquiring smaller, specialized shops focused on operationalization.

For example, DataRobot acquired data science operationalization startup ParallelM in June 2019 to extend its existing automated machine-learning-based data science platform into this critical arena.

Expect much more activity in this space to come. The time is right for data science platforms to embrace MLOps, to enable a successful machine learning model deployment phase at scale across their client organizations.


Want to keep up with machine learning trends?

Apply for free membership in the 451 Alliance, the premier IT think tank. Do I qualify?